The future of Augmented Reality | Interview with Real Wear CTO Timon Binder - The Generative AI Meetup Podcast

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(01:00:01):
All right, we're starting.
All right, welcome everybody to a special
episode this week. We have the CTO of a
really cool AR hardware company,
Realware. I have here Timon Binder, who
has been really interesting. We had an
amazing chat at our meetup this week.

(01:00:24):
Had lunch-- sorry, dinner at Google, and
wanted to continue the conversation with
our broader audience, and I have Mark
here on the line too.
So Timon is the CTO of Realware, and they
focus on the enterprise segment, building
headsets for AR applications in multiple

(01:00:48):
enterprise use cases.
And he has really interesting plans for
expanding to the consumer segment.(...)
With that, Timon, I'll let you introduce
yourself to our listeners and give them a
brief background about what made you
start this company
and your brief journey.
Yeah, thanks so much for introducing me.

(01:01:08):
It's been a pleasure. I mean, I also
would have never thought I would be here
now, meeting you just a few days ago.
It's been amazing. It's a
pleasure. Thank you so much.
And for context, you're here all the way
from Switzerland, just out here for a
couple of days,
shopping for a new location.
Exactly, as mentioned, so our
headquarters of the company is all the
way up in Vancouver State, Washington.
And so more and more being here, having

(01:01:29):
all the connection, it's more and more
apparent. We need to come here. So that's
why, as you mentioned, we're looking for
an office over here right now. So maybe
just very quickly to me, thanks so much
for the great introduction. As I am a
board member and CTO at Realware, for
everyone who knows me already or who
doesn't, I was previously the co-founder
and CTO at Almer, together with Sebastian

(01:01:50):
Beechan, who was our CEO and is currently
our CEO at Realware. So we started our
company in 2021, quite a lot of years
back. But it still went very fast, built
up the company with a very similar vision
to what Realware currently does, with the
focus of really trying to solve a end
customer problem all the way through. So

(01:02:10):
not only doing the hardware, but really
focusing on the whole solution that
actually makes something work. So we
didn't want to create just a hardware
product that people use. They run some
simple, funny apps on it, and then they
toss it away. But actually controlling
the whole process from the start to the
finish line and actually make it useful.
(...) And that's why we're dabbled more
and more into the also software aspect.

(01:02:31):
But then it got to such a point that we
were apparently so excited that we got
purchased or acquired by Realware,(...)
which was just by the end of last year.
And then a few months later, we got the
chance to take over the control of the
company and now drive our vision that we
have from the beginning
in this new form factor and have no

(01:02:52):
access to the US market also much more
since they are headquartered here.
Whereas all nervous previously
headquartered in Switzerland. That's also
where I still live, but more and more, I
have a strong feeling I'm going to be
here way, way more in the future.
So yeah, thank you again for having me.
Of course. Yeah, I mean, this is an
awesome place.(...) There's tons of
startups, tons of innovation.

(01:03:13):
And I think you, hopefully you had some
good conversations that are made up.
There's amazing, talented, accomplished,
intelligent people attending our events
every single week. And it's a hotbed of
innovation, not to mention funding
and a lot of customers too, hopefully.
(...) So on that note, can you tell us a

(01:03:33):
little bit more about who your customers
are? And from my conversation, my
recollection, it seems like right now
you're focused on the enterprise use
case. And that seems like where most of
the AR companies, apart from you are also
focused on. In contrast to some of the
consumer segments, like Meta and the
Apple Vision Pro, which is focused on

(01:03:55):
more consumer, either
gaming or entertainment.
So I think it's best to explain that a
little bit when we go into the use cases.
So what are the main use cases that
currently drive the B2B segment? What we
have identified, and we see that from our
customer basis is we can segment them
into two very big one. Most of them are
focused around manufacturing,(...) purely

(01:04:17):
in general. The very first one is
everything remote support. So if you have
any kind of a machine, any kind of
industrial equipment, or in general, any
equipment that is rather complex, they'll
be more complex than your average
printer. And you can't solve on your own,
you need to get help. And quite often
that's by someone who is not around, not
in the area, and so need to have a remote

(01:04:39):
call with someone who's not at location.
I mean, the remote worker takes a
headset, starts ideally as quickly and as
easy as possible, a
remote call with Zoom, WebEx,
Microsoft Teams, Google Meet. We would
have loved to do that,
currently doesn't work properly.
To really get that,(...) might be some

(01:04:59):
recommendation for that.(...) To really
get effortlessly a problem solved without
having two people in one location. And so
specifically during COVID, we saw a huge
spike on that use case, where people
really realized, oh, we are toast if we
can't go there and fix it. Specifically
if you can't even go into a country. So a
machine doesn't work in China, it's very
quickly, very expensive.(...) And that

(01:05:21):
also means in automotive industries, for
example, if you have a line that just
rolls through, if you have an issue, you
need to solve it ASAP. And you might not
have the expert at hand, so you might
need to do a call. And quite a lot, you
need to have video, you need to have your
hands free and really solve the issue. So
that's one very big use case focused
around anything, resolving
issues alone slash together.

(01:05:44):
The secondary one, which we see more and
more growing now also is logistics.
Very good example there is Coca Cola, for
example. So they use our headsets in
logistical warehouse picking. So there we
work very closely together with ISV's
company that build these software
applications end to end that actually
solve a picking problem.
(...) Now, what does that mean?

(01:06:06):
Picking means in big warehouses, I think
Mark, that might be interesting for you
as well. I think Amazon, they have that
as well. Even though they have a lot of
robotics built into that, if you have
larger items, there's a lot of, get this
item, bring it over here, stack it onto
this order and sort of getting everything
together. You walk around a lot, you have
forklifts, you drive on them and so on.

(01:06:27):
But you have a lot of use cases where you
need to go pick something,(...) put it
somewhere else, get the next thing, and
use the whole day, nothing else in that.
(...) And so there they wear
our devices in full shift,
usually eight hours, even in multi shifts
most of the time, where they have go
there, and they pair the devices with
scanners, which are Bluetooth scanners

(01:06:47):
from Zebra or Proglob, for example,
scan items, it automatically displays
where they need to go, how many they need
to pick. They also confirm orders through
voice, and it reduces their error rate by
double digit percentage and also
increases their efficiency. To the extent
that you can think about a worker, if you
take an average salary, getting so many

(01:07:07):
percentage more efficient, it gets very
quickly, very lucrative for the companies
to develop or put the system out that can
do such thing. And so that we see in
everything logistic, it's not only in
beverages, but everything that has large
scales, logistics that start to use more
and more of that. And it's very sticky,
because it's ERP integration, it's very

(01:07:28):
complex to get that going. And we can
only do that thanks to our great
partners, like TeamViewer, for example,
they built great software applications
that solve these issues.(...) So that is
the secondary, very big one. Maybe coming
back to the automotive one, the companies
we have there is Ford, for example, is a
very big one. A lot of other automotive
companies that work in this sectors,

(01:07:48):
large scale manufacturer, I sadly can't
mention everything, but everything that
produces big machines to solve issues of
ABB, everything, CNC machines, these big
expensive machines that are very complex
to entertain and resolve issues with.
(...) And then there's a third gap, I try
to keep it very short, so the medical
sector, we see coming more and more. The

(01:08:09):
more compact the devices gets, the more
we can tap into that market, and there it
is, more a sort of a, get an additional
doctor to watch over your shoulder(...)
and record sessions for purposes of
figuring out who is at fault if something
happens, but also just get a system or
get a lot of college people to watch into
an ongoing surgery that is maybe not so

(01:08:32):
often happening.(...) So if these are the
main focus we
currently see on the B2B side.
Yeah, that's really cool. When you think
about the medical use case, I'm sure you
run into a ton of compliance and I see
issues there, but coming back to some of
the industrial and enterprise use cases,
(...) so in our field, we're constantly

(01:08:52):
talking about the existential risk when
AI is gonna take our jobs. So it seems
like your product might also
eventually get replaced by AI.
Amazon definitely has robots that are
doing some kind of logistics work,
and with the growth of robotics and the
rapid advancements in the operating

(01:09:13):
system level, the training data where
they have simulated environments,
creating a bunch of use cases examples to
train these robots to operate in the real
world, it seems like it's only a matter
of time when you actually don't need
people roaming around, and maybe it might
be worth it to jump on the bandwagon

(01:09:34):
rather than fight against this current.
So on that note, I was kind of
brainstorming with you the idea of
potentially using some of these devices
that you have in the field, collect the
data, the visual data from all these
people working with these very
specialized niche use cases in logistics

(01:09:55):
or in automotive industries, or even
medicine, for example, use that as
training data to build models that can
eventually reach human level accuracy or
surpass that even. So any plans to do
that, either yourself or
partnering with other companies?
It's an amazing question actually,
because specifically in R&D, there's a

(01:10:16):
lot going on in that direction. And so
why is that the case? I
think as you mentioned,
the way we wear the devices or people
wear augmented reality devices, they are
first person view, perfect camera
footage, if you start actually trying to
capture data on how people resolve issue.
A lot of these robots, they need a lot of
training data to actually get things

(01:10:37):
done. But in the same time, these cameras
also if fed into AI models, very well
start to understand context.
Now, still now we all know the prices and
everything might not be ideal to feed
continuous streams into that, but the way
Moore's law goes, and we all know in what
direction that goes, it's probably only a

(01:10:57):
question of time.(...) And so we have a
lot of sensors placed directly like that
to take advantage of that.(...) Currently
they are not captured, also due to
privacy, it's need to be a little bit
more difficult, these kinds of things.
But more and more we have, if customer
opt into wanting to do something like
that, then we can start to do a lot with
these data. It's first person view, we
see exactly what you're doing. And that

(01:11:19):
allows for a lot of more efficient and
sort of new ways of interacting with the
world. So for example, very good examples
of logistics, there's still a lot of
interactions of the humans with, okay, I
picked four, I confirm, yes, and so on.
But we actually have everything to
confirm what exactly is going on. If you
look from a visual feed, okay, I see he
actually did pick the four pack of,(...)

(01:11:41):
let's say whatever to drink, so I can
automatically confirm it. And then in the
next step, I know if he places it wrong,
all without these confirmation steps. And
so we are researching very heavily into
that to take advantage of that. And for
that we place sensors into different
directions(...) to really capture
everything that is going on from a
contextual point of view, to really

(01:12:01):
understand what are we trying to solve?
What are we doing? What is the intent?
And what are we supposed to be doing?
(...) And then I guess that leads very
much into, yeah, what is the future gonna
hold? If everything is gonna be robot
only, which I believe it's a very long
term, probably gonna happen.
It still has a lot of niche points where
we need to sort of get advantage of

(01:12:22):
taking, taking advantage of these
technology and sort of become, I don't
wanna say better humans, sort of merge.
And I think Sebastian, my friend, he has
a better view on that, really merge with
the technology to get better at what we
are great at. And there's always gonna be
place for these kind of things. So I'm
not too much afraid of that, but I still
see that the challenges ahead might be,

(01:12:44):
that it might be a bit more difficult to
find a job, to just do picking in the
amount of people that
are currently doing that.
Yeah, I mean-- Yeah, the stuff is really
interesting. And I love the fact that you
are solving these problems in a way that
I think a lot of other people(...) don't
really think about, because I think a lot

(01:13:04):
of people are focused on solving things
only within the computer, but you're
actually out there working in the real
world, trying to solve real actual use
cases. So one thing I was curious about
with all of these sensors and talking
about AI models, are you finding that
you're able to run a lot of these AI

(01:13:25):
models directly on the hardware, or do
you need to hit some sort of cloud server
in order to do the processing? So for
example, if you are, I saw that you use,
I think Qualcomm chips in a lot of your
headsets, do you find that you're able to
run those on the edge and do the proper
local inference, or do you need to

(01:13:47):
actually hit some sort of cloud somewhere
in order to figure out if you are,
for whatever data
you're trying to process?
That's a very good question actually. And
we thought a lot about that because it's
a very strong sort of question on how you
actually wanna deploy and how you wanna
build the devices. But the issue is still
running now is, the technology is

(01:14:09):
advancing very rapidly. Nevertheless,
let's say the chips that aren't really
there yet to run everything locally yet.
And also there's two sort of approaches
to this. The very, very compact form
factors, they start to
offload a lot on the phones.
It's the easiest way you
have it in your pocket,
use it for connectivity as well. Still

(01:14:29):
run into the issue that sort of Apple
limits to some extent how you can
actually pair these devices.
Still is a little bit of mess.
But if you have that, then you can
offload some of your compute to your
phone. What we chose, and we come a
little bit more from the B2B side is we
have quite a good(...) sort of a little
bit more volume available. So we can do a

(01:14:49):
little bit more on device. So that's why
we don't actually require a phone.
Now, when you ask specific about all
these AI models, there at least currently
the way it is, it still runs in the
cloud. So it means you have still certain
features available if you're offline(...)
because you still need to navigate
through voice. Imagine you don't have an
internet connection, you
don't just wanna be stuck.

(01:15:10):
That means you're to very much simpler
ways of detecting voice. And usually that
means your default to sort of speech
commands that are free, I would say hard
coded, but they're sort of set in stone.
For example, if I wanna go back to the
home screen, I need to say, navigate
home. I wanna open the settings, it's
okay, open settings and so on.(...) That
is very much not intuitive and not

(01:15:30):
effortless. It's very bad.
It works, but specifically if you have
people and you wanna scale it to mass,
let's say markets in the end, you can't
have that. And so as soon as we hit an
internet connection, we can have the
advantage of everything running(...) very
capable. Now with that said, in the last
few months only, the models get better

(01:15:51):
and better and more efficient and more
efficient that we can start running the
first one on device,
but they're not yet deployed. I think
that's very important, but it's
definitely only a question about time
until it does, right? And I think on the
iPhone side, you can see it as well on
all the Google forms as well.(...) We see
pretty much that it gets easier and
easier and you can do more and more while

(01:16:11):
not having a connection.
What is very important to mention Mark
nevertheless is quite a lot of the use
cases that they actually have in the end
is that they have most of the time an
active connection ongoing.(...) So it
still might be or still is more favorable
to still do more and more in the cloud.
Yeah, use proper protocols to actually
share audio directly with the cloud and

(01:16:33):
get everything back. And also latency
wise, we got to a point that it's very,
very nice on what we see now. But yes, we
can definitely start doing more and more
and we have the chip set ready to the
extent that we can do so, yes.
Yeah, that's super interesting that
you're able to run it both locally and
then on the cloud. And I assume that once
you're running on the cloud, you're able

(01:16:54):
to get a lot more functionality.
So one other thing that I was kind of
curious is why you decided to go kind of
the voice assistant route as opposed to
using some other types of input modality.
Is this something that customers are
asking to say, hey, we wanna use

(01:17:14):
hands-free only? Because I know that
using other voice assistants, like a lot
of people, they don't actually want to
talk out loud, right? Because so for
example, at least not, maybe not in the
B2B space, but in the B2C space,
you might not wanna say like, Alexa, like
what's the weather? Like when you're on a
bus or something, right?

(01:17:36):
But if you maybe, I don't know, cause I
saw you were working with like Harley
Davidson and some of these companies,
like is this something that the companies
were asking for to say like, hey, we
really want a way to interact via voice
even if voice maybe has a higher error
rate? Cause I know that like if you have
voice, maybe there could be background

(01:17:57):
noise, people may mishear you, like it's
just gonna make mistakes, right? So I'm
just curious as to like, you know, why
you decided to go the voice route as
opposed to maybe having some sort of
physical device that you could kind of
type into or some other input modality.
I think that's a very good question,
right? Because we talked to Meta, also a
lot of others, there's different
approaches to this on the user

(01:18:19):
interactions of the whole UX, UI side. I
guess there's a lot of different
approaches to this. And I guess it's not
100% clear yet what will win the race
from hand tracking to, as you said,
voice, but also these different bands
that you can use to pinch and
controllers. And I guess the whole space
hasn't really converged on, this is the

(01:18:41):
right thing, at least not the way I see
it. But how did we get to thinking at
least that voice is a good way? So maybe
we need to start on what we think is the
end goal. The very much end goal we
believe is gonna be brain interface.(...)
Now, we're obviously not there yet, but a
lot of the device we have, they're very
close to sort of the area where you would
start to do something like that. And we

(01:19:01):
did invest some time into looking at
that, but it's obviously not yet the time
for that. So once you have brain
interface, then there you go, if you
think about it, it will happen. So we see
long-term sort of the star is that. Okay,
so in between,(...) what do we need to
solve something intuitive and effortless?
(...) We still have at least figured out,
and that's maybe more than a personal

(01:19:22):
approach, is people are very good at
explaining what they wanna do. And
specifically now, thanks to AI, it got
vastly better than it ever was before to
actually explain a task and then have it
done. So me personally, I'm already too
lazy to type stuff into OpenAI or
anything else. I just sit down,
specifically if I do business plan, or if

(01:19:43):
I do anything else more, let's say I plan
a new project, I should sit down, talk
for an hour, and it'd be like, just,
okay, write it down.(...) And it works
pretty good, I think already where it
goes now. And so from that part, we have
a strong belief that if properly
explained, you can
solve almost everything.
And with the import part, the hardest
point, and that's a very reasonable

(01:20:03):
question is, okay, how do you solve stuff
like, let's say I need to search for your
name, Mark, right? Your second name is
maybe a bit more difficult to pronounce,
or let's say an AI might not properly
write it. How do I do a call to you? Say
if I say, I wanna call Mark, I might have
10 people in my contacts that are called
Mark. But we figured out if we map it
properly, and we use stuff like fuzzy

(01:20:26):
search, if you know, sort of have a lot
of error rates built in and can still
reconfirm of the user, we can end up
calling the right people.(...) And so if
it goes very bad, you can do alpha
numeric, like alpha bravo Charlie Delta,
folks are not sure if you know that, sort
of feel like a fighter
chat pilot if you do that.
To actually really dictate something,

(01:20:47):
it's not ideal, but it works very well.
(...) And so the whole problem is, if you
argue on not using your phone, then you
wouldn't wanna have an additional device
that you need to charge, and be sure that
you have it with you, you just wanna have
one. And already one is too much, which
is also why we don't need the phone. So
we connect directly to wifi, you can

(01:21:07):
leave your home phone away. And
specifically for B2B, that is important,
they don't always have phones. They wanna
really make it on device, everything
only. For setup, it may be fine to use a
phone, but then you don't wanna touch it
anymore. So we needed to find this way to
really make it only like that. Then try
to keep it very brief, but another way is
obviously hand tracking. There the whole

(01:21:27):
compute side and power efficiencies,
which is not yet there. And on
Microsoft's HoloLens, Sebastian just
learned a lot that a lot just doesn't
work. And it's so often, you're just
fiddling around and it's just not easy
enough. So it's just why we end up
choosing completely to go voice only. And

(01:21:48):
specifically now, thanks to the
revolution going on, I think we were
quite lucky,(...) because I think it
could have not gone any better to hit
such a time where it actually gets so
much better because let's say four or
five years ago,(...) it was definitely
not as capable as it is today.
Yeah, I guess on that note, both Mark and
I used to work on voice assistant

(01:22:09):
features, Mark at Amazon, me at Google.
And we've seen how this older generation
of voice assistants,
these smart voice assistants,
as they were called pre-LOM,
were just massively overshadowed by the
ability of these LMs to truly understand
what you're saying. But then on the other

(01:22:30):
hand, they lacked all these integrations
and tools and APIs and all the features
that these old assistants were built
with.(...) So having
worked on Google Assistant
on these smart displays and smart
speakers for a while, it seems like
people resort to just a couple different

(01:22:51):
use cases pre-LOMs. They're like, "Imers
alarms,(...) I wanna wake up, I wanna
know when this task is done.
Calendar, productivity tools, help me
write a list of notes,
help me do a to-do list."
And then it was this open-ended question
of search and question answering and just

(01:23:11):
interacting with this device as an
intelligent assistant. And it seems like
we are really on the cusp of unlocking
this use case with voice assistants, just
asking it any kind of general question,
having it either pull from its own
internal knowledge from the neural
network or call a bunch of tools, go out

(01:23:33):
there, do a bunch of research, act as an
agent on your behalf, and then carry out
that task. So it seems like you have a
couple ideas brewing about exploring this
new frontier or this old and new frontier
combined.(...) And potentially entering
the consumer space. You want us to tell
us a little bit about that? Definitely.

(01:23:54):
So when we also, when we had the first
sort of LMs, so we had a closer look,
we're like, "Oh, okay, that's crazy."
Like the things that it can start doing,
I think a lot of people didn't really
realize. And we very quickly thought,
"Okay, what we are doing
is almost opposed to compared to what we
can do now, basically, using the
technology." And so where beforehand it

(01:24:16):
was not at all intuitive because they
needed to use predefined words and so on.
Now it could just talk freely. A very
simple example that I think I've
mentioned you as well is, imagine you
want to activate the flashlight. Somehow
I need to figure out how I go there. So
either I navigate into settings, which I
also don't know the word for, then I'm in
settings, need to scroll through multiple
pages, find the setting, and like, "Okay,

(01:24:36):
it's called light source." And sometimes
it's called flashlight, and sometimes
it's called lights and whatever. And then
I toggle it on, fine, it's all, I go back
to my application. And if you do voice
only for that, it's just a hassle.(...)
Whereas now the way it works, and that's
what we have now running on our operating
system is, I can just talk very freely. I
don't even need to mention flashlight. I

(01:24:57):
can just say, "Hey, it's very dark, I
need help." And then the way it works so
intuitively, it's like, "Oh, I have these
functions available. Where can I map it
to easiest? Oh, great,
is that total flashlight."
And when we first saw it, didn't even
take long, that was like two days. You
don't need to be a coding expert to get
that done, right? It's like,(...) "Whoa,
that is impressive." Like, "Okay, let's

(01:25:19):
put more functions. Like how many can we
do?" Now, obviously there's, in the
beginning context, we know it's only so
long, but if you just think a little bit
further, it's only a question about time
until longer and longer. And I think just
opening up just now with a million. So it
gets very quickly, very obvious that we
can start to do most of the things very
well, and it starts to behave very good.

(01:25:40):
They're still small bucks, but it really
starts to be able to do really what I
want. And so we see on the capabilities,
it's not only controlling the hardware,
which we can do very well since we
control the device itself, but really
sort of integrate with a variety of other
services that beforehand took immense
amount of time to get right and then
really interact with it. Obviously, still

(01:26:01):
limited to some extent to do the APIs
that are available, what it can do.(...)
But we also believe that in the end, if
you think it even further, that we can
send the agent out, doing tasks and
coming back, I guess more into the
direction of what operator does,(...)
what the ones who know what we're talking
about, sort of browser use, it's a new
ETH company that started to do similar

(01:26:23):
thing where you can actually analyze and
let multiple browsers actually start
doing stuff for you and just come back
with the results.(...) Where the
challenge is much more for us than how do
we actually represent information
properly on a limited display.
Whereas in the back, there might be a lot
of other things going on. And so we try
to take all these complex tasks and map

(01:26:44):
them on simplified UA elements and then
actually go out and make things really
intuitive and effortless.(...) And we see
that we can do that very well now with a
very limited amount of engineers. I think
whereas back in the days, that would have
been impossible to do.
Yeah, that's really interesting. So you

(01:27:05):
mentioned agents in your headsets. So in
my mind, headsets would be more for like
real time type feedback, whereas agents
might take 30 minutes to think about a
particular task. Is there a particular
use case that you found that agents are
useful with using like an

(01:27:28):
actual heads up type display?
It's a very good question. So we're not
100% sure yet what it will be. And I
think specifically on that aspect when it
comes to the agents,
all the ones I just mentioned with really
how to operate the device, call people
automatically, make notes, take picture
and send them through Microsoft Teams,
send emails and so on. They are very
obvious to us and we get a lot of pulls

(01:27:49):
from the customer basis. A lot of the
they wanna integrate with
these features, send emails,
send pictures that they take on the
device where beforehand they were just
stuck on the device,(...) send them out,
interact and really start to take proper
note of what the device can do. Now, when
we go more into the B2C segment, we see
more and more where we think, hey, these

(01:28:10):
agent will gonna be more and more
powerful to be used as well. And I guess
that's not really limited to AR devices
only because as much as you know, you can
start doing that also on a anywhere where
you have access, right? Let's say I wanna
have something to book me a flight, go
from A to B and I'm just on the run or
for us what we think about a lot in
Switzerland is we have a lot of trains.

(01:28:32):
So everyone needs to know when the next
train is leaving, which one I need to
take, where I need to go. So go out, get
me the best route and
come back once you're ready.
We don't have the proper pool yet because
we're not in B2C. So we're very cautious
to figuring out what the engineers think
is a great idea to what actually is a
good idea(...) because that is exactly

(01:28:53):
what we don't wanna do. Just have a lot
of guys in their basement thinking, oh my
gosh, that's great, let's do that, I
wanna use that. And then 90% is like, are
you nuts? No one uses that.
And so it's very hard to say what agent
exactly would be the ones used in the
end, but I almost think they will be
driven by what people use on the laptop
and on a phone first, I think, and then

(01:29:15):
the benefit is just gonna be, I will be
doing it on the run while I'm driving,
while I'm driving is maybe a little bit
the last thing you're gonna do, but like
while you're doing something else.(...)
So I think that would probably be the one
I could see most coming. And so for me as
a person, even though I don't think that
might be what others think, it's gonna be
a lot of that.(...) And the rest, and I
think that's what we talked about before

(01:29:36):
is more a lot of tasks which is get
emails properly done.
I'm so lazy, I ideally don't wanna write
my emails anymore. And I think we might
be the last area of generation that sort
of learn to type still properly.
Whereas it might be in the future more
and more voice-driven, you might even
start coding voice-driven, that's
obviously way more in the future, but

(01:29:58):
which is where at the hackathon in San
Francisco, where a lot of these concepts
start to become more and more realistic,
I think. So it's probably only a question
of what time until
when we start doing that.
Yeah, I mean, I'm personally sold on
voice being the dominant form of input. I
like this little extension(...) where I
can just hold the function key and it

(01:30:20):
listens and keeps typing,
not sponsored, whisper flow. But anyway,
you can find a bunch of these other tools
to allow you to just keep talking and
then convert that to text and then take
that text and then do something useful.
Coding is a popular use case because now
vibe coding has really taken off. So
being able to talk to cursor is way

(01:30:43):
faster than figuring out how to type the
exact phrasing that you want.(...) You
don't need to think about formatting, you
don't need to have the right kind of
syntax, just throw verbal spaghetti at
this LM and then they'll
just understand what happens.
But coming back to your point about
building a consumer,

(01:31:04):
I guess, voice assistants, there have
been a lot of companies that have tried
to do this. We talked about the Humane AI
Pin, which it got slammed by the media.
(...) MKVHD and a couple of people, I
think single-handedly destroyed that
entire company. And I don't think it was
his fault. I just don't think the product

(01:31:24):
figured out a good use case for
consumers.(...) I have the Rabbit R1
laying face down. Yeah, you told me you
need to show me. Just as a decoration
piece or something, because it's really
super orange. I was really fascinated by
the promise that they made because having
an agentic app store was incredible.(...)

(01:31:46):
And the fact that these things are kind
of hard to, it's hard to figure out use
cases, but then if you have an app store
with other average people building agents
with voice without much technical
ability, that would lead to an explosion
of useful features from
customers for customers.

(01:32:07):
But figuring out the ecosystem and what
will be the dominant player in the field,
what kind of form factor that will be,
whether it is AR glasses or phones. So
it's like, how do you go about figuring
out the product to build,(...) especially
coming from the enterprise segment? Is
there something on the consumer side that

(01:32:28):
you're focused on? That's a very good
question. Or maybe your own journey or
your own set of problems.
I would say in all of us, we have P2C to
some extent, because we think about our
own problem. And specifically from my
side, I never really worked in a big
company before. So we actually started
our company part of college. We didn't
even finish our master's degree. Went
back later out of force of family and

(01:32:52):
friends. We're like, you need to finish
that. So we finished it both Sebastian
and me exactly the same. So we thought a
lot about our own problem and how we
could use it. Now that's a bit difficult
in V2P because they have other problems.
And it took a while for us to really
understand their problems.(...) And in
the same realm, it was quite challenging,
but nevertheless very satisfying to then

(01:33:12):
start to understand what their problems
are. And we were very close to customer
for a very long period of time to truly
try to understand what are their problems
in a day to day business
environment. How can we help them?
And it's challenging to really get your
own feelings out of the way. And you
might think as an engineer, I really
think this feature is what they need, but

(01:33:33):
the end users really know much better
what they actually need. And so by
actually listening, we really tried to do
to wait for the pull and not push a
feature into it. Nevertheless, there's
obviously always certain thing. And I
think Apple probably did that best ever
to really decide to some extent the users
don't really know what they want. We're
just gonna force it upon them. And then
they realize, oh my gosh, that's what I

(01:33:53):
always wanted.(...) And so to some extent
we need to do that a little bit, but
really try to listen. And so the good
thing is now on building new features is
we have the real world has sold into
direction of a hundred thousand devices.
They went quite a few years back. So
they're not all really in use anymore,
but specifically the newer generations,

(01:34:14):
they are used quite regularly. Some of
them are almost used daily. So we have a
quite large customer group that actually
have an existing value that they see in
the product. And so they buy for that.
And then when it comes to our company, we
have a profitable part that is doing
that. Now we can build on top of that and
not just build on stories and nice to

(01:34:35):
have, but really try to give them more of
what they already love. And what we
realized more and more is if we go more
and more and listen, these problems
become more and more a B2C problem and
not only a B2B. So if we start, "Oh, I
wanna call this expert. I have an issue."
Very quickly after that, it's like,
"Okay, I wanna send him pictures of the

(01:34:56):
issue and messages." And then like, "Oh,
you wanna send messages?"(...) Starts to
get close to the B2C. Like, "Okay, I
wanna call this." Like, "Ah, could I make
a call through WhatsApp?"(...) Even
though as mentioned,
that's not yet possible.
It starts to drive more and more of these
use cases where we
start to realize, "Hey,
once we start running this on more
compact style of devices,(...) it

(01:35:18):
actually gets very useful more and more
also for the B2C segment. Whereas it's
not that we tried to really push
something and then see, "Let's try it
out." But it's more that we develop it
for B2B and more and more people are
like, "Oh, I started to
do my emails on here."
Whereas it might've not been intended for
that, but then more and more devices join
the party and actually get this use case

(01:35:40):
more and more valuable. And so that's
what we try to do. We really try to
listen and get our own opinion out. Even
though I still think we need to mail this
email thing, and that's
really my personal view,
just because I could see myself, if I
have a really long email, I'm like, "Ah,
can I need to read it? I can let it
summarize, but then I put down." Whereas
if I have someone reading to me, I really

(01:36:01):
do the reply properly drafts, I can
refine the draft and just send it out.
That's what I wish in dream. And I really
can see it's probably only a question
about orders and not years anymore.
It's funny that you mentioned email
because I feel like everyone has maybe
two or three core ideas that they keep
trying to chase with every new wave of

(01:36:22):
technology. When we have the dot com
boom, when we had this new on demand
explosion, now when we have LLMs, people
try to build tools for email, they build
tools for events, and they build tools
for networking. And I think Notion just
came out with another email client, which
is supercharged with AI and helped you
navigate it without as much tedium.(...)

(01:36:45):
There was like superhuman email. I think
they also tried to use a lot more AI. I
don't think they've ever come up with a
public product. It's always been in
closed beta for the last
five years or something.
And that's also very important. We
haven't solved that. We can't really do
emails, but the way I would see myself

(01:37:07):
using it daily, we're not yet there. But
I can see sort of one iteration after the
other is like, oh, we're definitely
getting there. It's really just a
question on when not if.
Yeah, I think, yeah. I guess my takeaway
from this meta analysis is that I think
people have a lot of problems that
they're trying to solve, and they keep
hoping that this next wave of technology
will be the one to solve all their

(01:37:28):
issues. And I guess, how do you like
solve for the present versus solve for
the future when a lot of possibilities
will be unlocked? And do you have like a
current three to six month plan and like
a three to five year plan that are
working in parallel but separate?

(01:37:48):
Yeah, definitely. So what we do,(...) and
so we have a lot of strategic alignments
that are not only on the management, but
we try to get everyone involved. And so,
right, our company is still rather
smallish. I have around 50 people. It's
not thousands.(...) So we can get that
going much easier and really
start to think of what does
the majority of us actually think is a

(01:38:09):
good idea. Not just me, not just our CEO,
just what does everyone actually think is
a good idea? And then if you think, oh,
we can do that in this sort of time
properly, then let's maybe do a little
bit less, but let's do it properly.
Because we are deploying on B2B solutions
in general. It's very important that it
actually runs because if a warehouse goes
down, yeah, you don't wanna be caught.

(01:38:31):
It's not fun. So you need to take care
that you do one small step after the
other and actually solve something. And
we sometimes have this issue that we feel
the urgency a little bit too much. We're
like, oh, we're too slow. Specifically,
we're sort of a startup.
We wanna be fast, agile.
But it helped a lot by really making what
we're doing properly and only add one
more step if we actually think it's ready

(01:38:53):
for the next step. And that means if we
lay it out in phases, we have it
currently in sort of three phases on the
parts of what are the main capabilities
that are really needed to get an
intuitive and effortless usage of the
device when it comes to hardware
controls. So that's the first step.
And that is basically complete now.(...)
It's rolled out on a lot of our

(01:39:13):
customers, but large scale only roll out
is gonna be in a few weeks. No, that's
the first step. And the second step is,
okay, what are these sort of intermediate
tasks that now add a lot of value, but
are maybe not yet to replying to emails
and solving all of these issues, but
really making a direct call with someone
I wanna have,(...) really putting notes

(01:39:34):
down, sending pictures to the right
person properly and annotating them,
adding short text and send them back and
forth.(...) And so that is gonna be
solved in probably three to five months
after that. Also rather quickly with the
possibility to it taking maybe two, three
months longer, depends on how many
developers we can get on it.

(01:39:55):
And then from there, it's sort of where
phase three starts where I guess the fun
goes. And it doesn't mean we can
beforehand already do some of these
tasks, because that usually works, but it
wouldn't be enough that we
would say the feature is ready.
Where we really think sort of long style
messaging back and forth with proper
rewriting that you actually have what you
want, that you start to go into proper

(01:40:16):
email replies and so on. And then in this
phase, we also have sort of calendar
scheduling, which also is part in phase
two, the way we have it now, but really
get it more detailed. Show me that, show
me that. And having a variety of other
use cases mapped on these basic UI
elements. And so for that, we built a lot
of core elements that can map a lot of
different information accessible to API.

(01:40:39):
So for example, if I wanna reply my
team's messages, that might be different
on what is running in the background, but
what is represented to the user, it is
very much exactly the same. So I'm still
representing UI elements. I just fill
them with different content. And so it
scales much more to then
whatever integrations we have. And we had

(01:41:00):
a quick talk about that, right? It's in
the end only limited by what you can
integrate with.(...) But also in the end,
what do you actually have that actually
solve something? Because if you have it
just, I'd integrate, you can do it, but
it doesn't actually work.
Then it's not gonna be used.(...) And
that is sort of the plan moving forward.
And then even further ahead,(...) which

(01:41:21):
we also still start now because it's more
on the hardware side,(...) and we have
more and more of these contextual
awareness(...) sensors feeding back into
the AI and starting to harden it. It goes
very much into what Hasla does and others
when they have their cars driving and
someone sort of takes over the steering
wheel, it's like, whoa, almost an
accident. It's like, okay, let's annotate

(01:41:42):
that. Let's figure out what's going wrong
and let's train it and make it better.
Whereas we have that as well, because we
have the context, we have a lot of
information about the human using it. So
we can feed that back in and make it
better.(...) But that's, I guess then
very realistically seen that's more next
year than this year.

(01:42:03):
Yeah, that's super interesting. So one
thing I was kind of curious about is when
it comes to kind of knowing, kind of
going on top of what Shashank was kind of
asking to think about what was gonna be
in the future. But I kind of want to go
back into the past a little bit and talk

(01:42:24):
about your first customers and how did
you kind of sell people on the idea that
to say like, hey, we want AR headsets,
right? Cause AR headsets are kind of,
like I think they're kind of new
technology, relatively modern, whatnot.
And how did you say like, hey, we have
this product, we have this AR headset,

(01:42:45):
like this is something that you need or
did they come to you to say like, hey, we
have these problems, we wanna solve like
X, Y and Z things. How did those
conversations go initially to say like,
hey, we have this AR headset or did
people come to you say like, hey, can you
build us this? This is like a type of use
case that we're trying to solve. Like I

(01:43:07):
think that is something that a lot of
people who are starting new companies may
find useful to know how you kind of
reached out to the customers,
like how you listened to them, like what
types of steps did you do to figure out
like actually they're the things that
they wanted to solve.
Yeah, that's very interesting. So what we
have or how it went for us is we started

(01:43:27):
prototyping a lot, right? We didn't
really have an idea yet. Specifically, if
you start from zero, you have a lot of
chicken and egg problems. So do we wanna
put the optical system to the top, the
bottom, wanna come from the side? It's
like, you can compromise almost every
aspect and if it's almost an end
dimensional mat,(...) with different
things you can compromise on, we really

(01:43:49):
tried to find a sweet spot on what you
want. And once you have that built sort
of fish, we had some articles locally in
Switzerland that came out, sort of talked
about the vision, which made a lot of
companies very intrigued. And be like,
hey, I wanna try that. And interesting
enough, these were the companies which
had the biggest pain point that we tried

(01:44:11):
to solve at that point. And that was
getting people together when
you can't get them together.
It was one big one. And then the second
was during COVID.(...) I was actually
slightly pre-COVID, but it was still for
some company, it was a bigger issue
already,(...) not otherwise. But yeah, it
was, I guess you could all say it's sort
of around the same time.(...) But you

(01:44:32):
then have someone who is not in location.
So we had a company called Löpfe.(...)
It's also very open day. They were our
first customer,(...) together with
another customer. They had a lot of GARN
manufacturing in India, kind of.(...) So
they really couldn't be in location. It's
very hassling environment. It's very

(01:44:52):
industrial. It's humid. It's hot.
Phone's overheat. And so our product was
very well optimized for sort of hitting
that sweet spot. And let's do a remote
call with someone who's
there, not in Switzerland,
with someone who actually is in
Switzerland. And so they actually reached
out to us and they helped us actually
harden the product because we actually,

(01:45:14):
the very first time we met them, we
didn't really have something running. It
was more like, Löpfe sort of put it on
the Raspberry Pi, still somewhere else.
And we had helmets in the beginning. We
didn't know what form factor. It was just
big, huge, clunky thing. And over time,
and thanks to listening to them, we sort
of got it to where it ended up being.
(...) And so that is how we got the first
ones. And then it's still, and I guess

(01:45:37):
that's also a good thing, but at the same
time a bad thing. It has a very strong
innovative approach. Sort of be this, ah,
AR, great, I wanna try this. So a lot of
innovation department tend to be like,
outreach, and be like, oh great, I'm
gonna use this for, who knows how many
use cases, which at that time and still
now even are still not possible.(...) But

(01:45:59):
they needed to have a story on how we're
innovative. Yeah, we have this new
innovative product. And more and more we
tended to actually not sell to them
because we realized they're just laying
around. They don't actually have a use
case. They just wanna play around with
it. And so even though we get a lot of
coverage also news wise and people
actually are interested on the innovation
side, they ended up never being the ones
who actually ended up using it. Because

(01:46:21):
they were dreaming and they helped us
maybe putting a direction, but they never
drove a use case to the end. So I think
that's how it was in the beginning.
Yeah, that's super interesting. And I
love to hear how things kind of grew. And
I imagine that once people actually were
giving you money to solve the use case,
that's a pretty strong indicator to say,

(01:46:42):
"Hey, this is actually a problem we're
solving." Not just like you
brainstorming. Because I mean, like, you
might talk to us and we have all kinds of
ideas, but we're not actually handing you
a paycheck to say, "Hey, like here's a
hundred thousand dollars, go build this."
But if somebody says like, "Hey, we have
this particular problem. Here's money to
solve the problem." That's a really

(01:47:03):
strong indicator. They actually really
truly care about that particular problem.
So actually one other thing I kind of
wanted to get into, it's not like exactly
an AI question, but you mentioned that in
your initial kind of prototypes, you were
using a Raspberry Pi, which is using
Linux, Ubuntu. But right now you are

(01:47:25):
using mostly Android and then you are
using your new OS, which is the Alma OS.
And I guess we were talking before this,
it'd be now rebranded to REOS. But I'm
curious as to why you decided to kind of
go the Android route as
opposed to going the Linux route.
Very good question. And I already

(01:47:45):
mentioned it beforehand, that was a long
discussion. And we had experts from all
sides in the beginning. I mean, at least
as many as we had access to, given that
we were not here in this area, might've
been better to be here for that.
To really think about, "Hey, what do you
wanna do?" And obviously starting on
Raspberry Pi, we were coming from a
different direction. That was really much
fast prototyping because no one of us

(01:48:06):
actually had any clue of AOSP.(...) So
it's just easy to take Raspberry Pi. We
had in just simple slide shows and we
didn't even have voice commands going. In
the beginning, it was literally just
slide shows. Someone had a Bluetooth
keyboard sitting two meters away. If
someone said, "Ah, next step, just hit
the button." Like, "I don't need to build
voice commands to see if it's actually
useful that way." We did tests like that

(01:48:28):
in the beginning only. And they were
aware, obviously, we didn't tell them
we're faking it. They knew we were faking
it, but just to see, "Hey, would it
actually work? Could I, would that make
sense to me? Just use my voice," even
though we didn't have it done. So we
believed in fast iterating, like as fast
as only possible. No one is gonna write
any code to get anything and just get the
Raspberry up somehow. We even had at one

(01:48:50):
point, we had just phones and video calls
with Android phones with USB-C to add
other displays or secondary screens, put
that in front, just read, "Harker" on
that side. But then as you asked, right,
at some point we need to decide what
route we're going.(...) And on the power
optimization-wise, everyone was like,
"Okay, definitely not Android."
Though we thought, "Okay, it's a good

(01:49:10):
point." But then we knew that a lot of
apps on the whole ecosystem still is
building on top of Android. So if we
wanna use and take these capabilities,
we might as well just go that direction.
And that was actually the main reason why
we ended up deciding, "Let's go with
Android and try to throw everything out
that we're not needing."(...) That's also
what is currently the case. So it really

(01:49:32):
goes from AOSP side, and it's like,
"Okay, we don't need this. Why? We don't
need this. Goodbye." Still is not as
power efficient, clearly not as it could
be. But it means out of the box, 90% of
all applications basically run. Doesn't
mean you can navigate them well because
the UI is clearly not optimized, but I
couldn't put an app like Twitch, for

(01:49:53):
example, on it. I just, I can do a Twitch
call because it runs in proper Android.
It's a mess to go and change my Wi-Fi
settings, for example, but it works.
And so from there, the main challenge
then was, okay, now how do we actually
get the interaction going, given that
it's not a touch-optimized device? But in
the end, coming back to your question, it

(01:50:13):
was very valuable because we saw a lot of
competitors like RealWear now doing the
same thing as us. And so therefore we
could leverage their ecosystem as well to
get them also to port over onto our
devices without needing to modify
anything. Now, there still was
modification needed on the UI side.
There's different aspect ratios and

(01:50:33):
cameras are different and so on, but it's
much quicker than if we would need to
have everything to run on a non-Android
device. That is the main reason on why we
went that route. And also because the
form factor itself always allows to be a
little bit less power efficient to what I
guess the direction is, if you really
wanna go to these type of classes that

(01:50:54):
you are wearing work, but it's very, very
difficult. You cannot waste one single
milliamp somewhere. And so it really
needs to be super efficient.
Yeah, it makes sense. And I tend to
agree. I think that Android does make a
lot of sense and it seems like a lot of

(01:51:14):
companies are kind of going that way. So
for example, I know the Oculus Quest is
going the full Android route. A lot of
companies are going all in on Android and
you can also leverage a lot of people's
experience developing Android. Cause I
know there's like plenty of people who
have lots of experience developing in
Android working on these types of

(01:51:35):
particular use cases. So I'm sure like
the developer pool is potentially bigger
than if you cooked up your own completely
new OS(...) on top of a boot or whatever,
right? From the ground up. I mean, you
might have a little bit more power
control, but I mean, even on Android, you
can have some power control as well. So,
I mean, there's some really, really
efficient devices that are running on

(01:51:57):
pure Android. So I mean, I think it makes
sense, right? It makes a lot of sense.
So kind of switching the
topic a little bit as well, right?
I know that you're developing a lot of
the hardware and you are, I think, I
don't know exactly where the headsets are
being assembled, but how have tariffs

(01:52:19):
kind of impacted everything? So I know
it's kind of a topical thing now with,
you know, all of the Trump administration
putting everything in and is that
affecting you? Is it like negative? I
don't know how that's all going.
Sure.
So it's actually quite funny, right? So
we are obviously not born here in the US,

(01:52:41):
right? So we're from Switzerland. And I
guess all of Europe looks a little bit
different at the US with regards to that.
Most might not really care. They're like,
"Oh, they're having fun over here."
Obviously it touched us more, obviously.
So we have manufacturing in China, and
currently specifically assembly is over
there. So we source a lot of active
components, not from China, right?
Qualcomm, Snapback and chipsets and so
on. They're obviously not coming from

(01:53:03):
there,(...) but they're from the region
at least. And so China has this advantage
of really having all these components in
one place, right? To build these compact
products, you need a lot of know-how, but
you also need all the suppliers in one
region. You need to be fast, like so
Shenzhen and those Shanghai, these areas,
they have everything in one location.
It's like you have shopping malls. You

(01:53:24):
just go, "I pick my camera, I pick my
sensor here, okay. Chipsets, get it
together, get it done. Two weeks
prototype done." Whereas you just can't
do that in these areas here. Now that
means taking advantage of that and also
of the cost implication,(...) we have a
lot of all devices manufactured there.
Okay, now what does that mean in terms of
terrace? When it got announced, we

(01:53:46):
didn't, let's say, immediately ship full
planes over to the US to escape the
terrace like Apple did or some of the
pharmaceutical companies as well did.
Because we knew in the long run,
specifically on the electrical side, that
cannot be permanent. I guess it's almost
impossible. Like if you think about the

(01:54:07):
iPhone, for example, where 90% of the
iPhone are still produced
in China, some part in India.
If you have, I should say, 50% perif,
Apple is running good margins, but still
there's only so many margins. And that
costs very fast, really deep. So we
always, maybe we were a little bit naive
to think that I cannot be too long. So we

(01:54:29):
thought it's just a power play. Someone,
who is gonna hold this up long enough
until actually having a resolution? And
so right before our day announced that it
got exempt, that electrical stuff is
exempt,(...) at least for now,(...) we
were already starting to thinking, okay,
maybe we need to start thinking about the
strategy and what we can do.(...) But we
ended up now at least being in a position

(01:54:50):
that it's not a big problem. Now, why
exactly is it not a big problem for us?
Now, what we have is in B2B, the products
tend to be more expensive.
And the bill of material tends not
necessarily be more expensive. If you
look at our products, that's why I can't
disclose the details obviously, but it's
chip sets, a little bit of electronics,

(01:55:11):
and a little bit more plastic than other
things which are more compact. This thing
itself doesn't make it much more
expensive.(...) So without saying the
details, what you can automatically infer
from that is the margin's largest much
better. And if you do very simple
calculations, let's assume we have a
product that sells for 10,000 and it
costs you a thousand to produce. That

(01:55:32):
tariff actually is only made on the bill
of material.(...) And so let's say even
you have 100%, you double the price, then
we're going from 1,000 to 2,000. And so
if I forward that all to the customer,
it will mean it will only need to
increase the price by 10%, which really
doesn't affect you that much. And that's
a very hardcore example.

(01:55:53):
But if you're running a product that you
produce at 400 and you're selling at 500,
(...) and now you're having 100, yeah, I
mean, you're screwed.
And so that's the reason, thanks to our,
in short, good margins,
it didn't really affect us. Even worst
case scenario would not affect us really,
but it would definitely affect Apple or

(01:56:13):
anyone else sending consumer products.
And to the extent that we believe that
cannot be held up for more than like
three months, then there's gonna be a
huge amount of other problems around. So
that's why I think we had all the fun
around it. And what else was good is we
had enough stock in the US. We could have
lived for quite a long time until we
actually would need to import and
actually pay the tariffs.

(01:56:37):
Yeah, I guess I kind of understand from
working on the hardware org for a couple
of years now, looking at some of the bill
of materials breakdown, it's like so
minuscule. Some of these components are
worth pennies. And it makes sense for
hardware manufacturers because at scale,
these pennies add up. But if you're

(01:56:59):
working on relatively small volume and
charging customers primarily for the cost
of stock, for the cost of services, and
the AI and the support and all the
software that you built around the
hardware, I guess maybe that makes sense
where you're not affected as much.
Yeah, definitely. And I think it really
also depends on what country is good.
Now, I don't wanna go too political

(01:57:20):
obviously, but trade in this sense is if
certain countries are very good at
certain things, might somehow need to be
able to leverage it. And I think
Switzerland is maybe a very extreme
example of that, how we grew up really,
because we don't have a lot of resources.
You sort of need to share what is around.
And so I think it's obviously very
difficult to force everything into the US

(01:57:41):
through that. Nevertheless, I see, let's
say, the general belief behind it. And I
still think, and I think that's actually
quite sad if technology and know-how gets
away and is not around anymore,
specifically in the Western hemisphere.
And so I think that is a very good drive
to get that. To what extent that can be

(01:58:02):
played or negotiated or power played
pushed through, it's probably a question.
(...) But I still think it will be very
dangerous to lose that know-how, I think.
And so I'm very much in favor for all of
that. Whereas I still think, obviously,
if you do large scale manufacturing at
millions of devices,
you're probably gonna be very difficult
off if you would say you couldn't do the

(01:58:22):
iPhone and the US would be impossible. I
mean, you could, but no, I could really
afford it anymore. So it's sort of a
trade-off, I think, between how low do
you wanna have prices to go and what
other sort of aspects
are you driving, I think.
Yeah, that makes sense.
And it's interesting what you say about

(01:58:44):
Shenzhen, to say, oh, there's these
markets here that you can easily make
prototypes. Actually, I remember there
was a YouTube video that I saw a few
years ago. This was back when the iPhone
7 first came out when they didn't have a
headphone jack. And then there was a guy
who went around to the markets in
Shenzhen, and he just was able to
actually just create an iPhone, just from

(01:59:05):
parts bins, and he was able to actually
create one with the headphone jack. So
that made me always want to go visit
Shenzhen. It seems like this kind of
really awesome place,
because I've never been there, but I
really wanna go check it out.
So when you mentioned that manufacturing
in China is just more possible,(...) how

(01:59:27):
difficult do you think it would be to
manufacture something that you create
in the United States if you even wanted
to go that route? Is that even possible?
Or do you think that it's kind of like a
pipe dream of the administration to say,
hey, this is something that you'd be able
to do, just like me? Because in my mind,

(01:59:48):
it would be incredibly difficult to just
drop all the manufacturing in China, and
then just go spin up some sort of
manufacturing plant in the United States.
Because I don't know if
the know-how is there,
there's enough people who would even,
like if the materials are there, like the
whole supply chain needs to be built up.
So I'm just curious, like if you have a

(02:00:09):
thought, like is this even like possible
that like could you even start like
tomorrow, even if you really wanted to?
It's a very good question because we have
to still look around on, let's say for
example, assembly,(...) get them
assembled somewhere else. I think that's
the first step you look at. And that
obviously is very easy to do, right?(...)
Labor cost is higher, but you can
assemble it.(...) Now testing equipment

(02:00:31):
after assembly gets again a little bit
more difficult, but you can also sort of
scrap them together and you can do that
as well. Now that makes your device
assembled in the US, not yet made in the
US necessarily. And now different
countries have different laws on what
actually needs to happen such that the
product is allowed to be claimed made in
and not just assembled in, which then

(02:00:52):
makes a big difference when you go to
other discussions again.
But from that point of view, actually
wouldn't want to make everything in house
in the US, let's say as an example.
There's certain things which are easier,
plastics for example, plastic injection,
if not maybe as much know how around as
in China, clearly not even close, you can
do it. Same in Switzerland, there's quite

(02:01:14):
some companies, large scale that do
injection molding in house. You pay way
more for the molds and then it gets an
issue if you prototype, right? You need
to modify molds and the mold needs to be
very precise, specifically
if you have a steel molds.
It takes quite a lot of time, five, six
weeks just to adapt the mold and cost
tens of thousands just to make a mold.

(02:01:36):
And for each part you need to have a
mold. So it gets very costly if you just
want to get started, specifically on
startup side. But let's assume you
already have everything, you can move the
molds that you can do and there is
machines around. So you can do the
plastic, that's not an issue.(...) I
guess the electronics part is the
hardest. You can have assembly, assembly,
you can get that done as well. That's
also done in Switzerland.(...) I guess

(02:01:58):
the hardest point I would see from the
capabilities which is
hard to find is flex print.
I know it's possible, but we had very
difficulties to find that anywhere else
than in China. And even there you want to
be fast and iterate. We have flexible PCB
and it's just literally unmatched. I
guess you could probably build it out,
but then again, again, the cost increase

(02:02:18):
will be dramatic.(...) And I think where
it really comes down to is then this
integration of all of these different
parts. I need to have this and then I
need to have the experts on that quickly
integrated with that. And then I quickly
move the parts from here over here. I
quickly get the new camera sensors and
next day I have it in house because it's
just my neighborhood as it. I think that
is the hardest part to get core term. I

(02:02:40):
still think you could get it long-term,
but I guess that's very hard
to get just instantly over.
So I think that is the biggest challenge.
And then the second probably is then
maybe a little bit work ethics,
where let's say if you have an issue in
China, then everyone is not going away on

(02:03:02):
the weekend. You're gonna solve it. This
work attitude just doesn't really exist
to this extent in Europe at least. In the
US probably also not as much.
There is not as much vacation. They
really have a really strong focus on the
company and the vision and wanting to do
that. And obviously it comes with
negative, a lot of drawbacks

(02:03:23):
on what that actually means.
At least everything I saw on C is,
let's say the work environments are okay.
They're definitely not as nice as on our
desks. Well, I think there's a big
difference, but you need to have a
culture as well, which is ready to solve
no matter what.(...) And actually we just
had a discussion with an investor this
morning, who also said it's so difficult.

(02:03:45):
They tried to manufacture certain very
complex part they really wanted to. They
invested a lot of money to get it done
here, add all the stuff, but the people
just didn't have enough dedication to
solve it. So they needed to still have a
lot of external people then flown in who
actually solve it in the end. And then at
the end, they need to say, hey, we just
can't do it. We need to move it out. So

(02:04:07):
do I think it's possible? Yeah,
definitely. Is it challenging? Definitely
as well. And so I think it still needs to
be kickstart. And I think it can be a
good kickstart to get specifically the
things that are doable back.
And so, yeah, I really hope that will be
possible to as much as possible without
losing the advantage of
the lower price, I would say.

(02:04:28):
Yeah, that's really interesting.
And it's interesting what you say about
the culture fit, because it's not
something that I actually consider, but
that's true. I think that the Western
style of work versus, I guess the Eastern
style of work is definitely a lot
different. So for example, I've lived and

(02:04:50):
worked in Japan for years. And I think
that in Japan, there was a high,
I think people just take their job
incredibly seriously here, right? So
people will work late into the nights.
(...) People will maybe not be the most
efficient, but if your boss is in the
office, you will be in the office and you

(02:05:12):
will stay there working hard. And I think
that people here are incredibly good at
following process. And thus,(...) I can't
speak as much for China, but at least I
can say in Japan, it allows you to get
just incredibly high quality goods. And
then also, you can also rely on the

(02:05:33):
worker because I think at least in Japan,
you have a lot of companies where you're
able to work there almost your entire
career if you want to. So at the US or in
other countries, I don't know what it is
in Switzerland, a lot of people will job
hop every few years. But I think in Asian

(02:05:53):
countries, it's much more common where
people will kind of look at these types
of jobs as a career where you will go
join the company at 22, as soon as you
graduate from college, and then you will
stay there until you're 65 and retire and
then maybe start working a little bit
later, but you're there and you are
dedicated to the company. So I think that

(02:06:15):
kind of mindset just doesn't exist in the
West. And I don't know if that
necessarily translates to better
products, but I do think that it is like
a distinct difference and that probably
maybe does translate to better products.
So yeah, it's
interesting how you say that.

(02:06:36):
And I don't feel right,
right? This is only what I see.
There's a lot of great advantages to both
sides, right? I think that is very
important to say all the credible people
we know from China, all around Asia,
Europe and the US, the greatest thing
usually happen if the greatest part of
everything sort of come together and
build something great together.

(02:06:58):
Yeah, for sure, for sure. And I think
that you can't have one without the
other, right?(...) I think that on the
one hand, it's good to be kind of
dedicated to a problem, but then on the
other hand,(...) it's good to have kind
of like the free thinker who maybe job
hops every now and then because maybe
they'll be more willing to speak their
minds and they'd be more okay.

(02:07:19):
More insights from different thing. It
might help a lot of innovating because I
think that's another point where I think
(...) the US is just so many steps ahead.
Also specifically if I compare
Switzerland now,(...) just to think
outside of the box and like really try to
do more and really think further ahead
than just a little bit. I think
specifically in Switzerland, I think

(02:07:40):
that's a little bit hard for us sometimes
to see that people, you almost need to
think small. You have a very small
market. You only have 8 million people.
Then we speak four different languages in
the small area. It's like, it's not
really the place to scale. So you have
other things that evolve and I think that
very flourish there or maybe not huge

(02:08:01):
versions, very built automotive companies
or go to the moon yet, I think, but it's
different I think, yeah.
On that note, maybe since we were talking
about work culture and we
talked about your background,
kind of taking a break from your masters
and starting a company,(...) tinkering on
projects on the side, eventually building

(02:08:23):
something that lasts out of that. Do you
have any advice for our listeners who may
be still in college or who might be like
itching to start something else? I would
think get their hands on some of these
hardware products, tinker with and build
applications in VR, AR,
work with hardware too, in addition to

(02:08:44):
software.(...) Any tips, guidance,
especially like today,(...) because we
have a lot of tools that make it a lot
easier to get up and running. And maybe
like, if you could suggest things that
they could do, maybe visit China, go to
Shenzhen, pick up a couple of pieces and
like start soldering them together.

(02:09:04):
Yeah, definitely. So what I would say
very much in the beginning first is, in
general, if you start any kind of
endeavor, and that's maybe more personal
advice on that specifically,
look for a friend who joins you on the
journey. Because so I did that together
with Sebastian and he even has the idea
slightly before me even, this is very

(02:09:25):
important, you can't do it alone if you
wanna go all the way. I think we had
always us together and it's very nice to
have someone sparing partner to really
exchange ideas and figure out how crazy
you are to actually solve something that
goes with you through all the good times
and all the bad times and in all sense
and get a cool team, fun team around that
it doesn't feel like work because you're

(02:09:46):
working a lot and you're doing a lot of
hours. And if you feel like working, it's
really not fun in the end. Whereas if you
love what you're doing, it really doesn't
feel like work because that's more a very
general approach. But coming specifically
on the hardware side,(...) I'm obviously
an engineer because I have an engineering
background and that obviously helps. If
you have some ideas, I guess it's very
difficult to join the hardware game if

(02:10:07):
you have no idea what a pull up resistor
is or just don't really know all of these
aspects. But assuming you have an
engineering background to some extent, I
think obviously Arduino is sort of the
first platform that people would
recommend. That's probably more high
school, say rather than you obviously
jump on Raspberry Pi, you
can do a lot of these things.
What helped us a lot is really think of a

(02:10:29):
problem and then trying to solve it, no
matter what. And it's very important that
you usually also don't need to go down a
specific road of do I wanna choose
Arduino? Do I wanna choose Raspberry Pi?
What do I wanna do? Because there's a lot
of platform that can solve the issue. But
if you could just get started on
something, get to know what it works, the
in and out, and what it can control with,
you can get a long, long way. But then as

(02:10:52):
you said, I think going to China would
definitely help. And so for example,
Sebastian, the other co-founder, our CEO,
he was actually studying in China. I
think that helped a lot. Specifically in
the beginning, he had that whole drive
of, oh my gosh, they are crazy over
there, crazy fast. They're saying
Shenzhen speed, it's there so much
faster, we need to take this mindset and

(02:11:14):
get things done. And specifically if
you're a little bit in China, if you have
the opportunity to be there, it's great
to see what you can actually do and how
fast you can do things.(...) Where it
seems sometimes it takes a very long
time. And why that actually is again is
because you have everything together. And
everyone knows where is what.(...) And I
think Marco, you mentioned is a very good

(02:11:35):
example, adding an audio jack to an
iPhone. I think I also saw videos where
they did USB-C to previous iPhone.
Because they didn't have, I mean, they
only had lightning before obviously. And
these kind of things you can get done
with brilliant minds altogether that
understand the technology. And I think
specifically now that you can learn so
fast, specifically with

(02:11:56):
all the LLMs coming up,
a new subject can help you solving it. It
has never been that great. And obviously
it's easier in software, but it's also a
complete new area, I think also for
hardware, to learn things, get things
done. Because even if you're doing
hardware, you're touching software all
the time. Because your hardware isn't
going to do anything by just soldering it

(02:12:18):
together. Except if
you're doing a light switch.
And so I think if you get onto a platform
and you really get good at it, to the
extent that you can do something with it,
you really quickly start to understand
what else you can do with it. And the
ball just keeps rolling until suddenly
you're running on dev kits of complete
other chipsets and running very more

(02:12:39):
complex things down the road.
But obviously also for the ones who don't
study, I would still say studying helps a
lot, but it might not
be a requirement anymore.
I don't want to open that kind of worms,
but you can learn a lot
without going to study as well.
Yeah, that makes sense. I like the advice

(02:13:02):
to start with potentially an Arduino or a
Raspberry Pi. I think, you know, I've
played around with that stuff in college
and it was pretty useful. But I agree
with you. I think that just going and
trying to solve a problem,(...) it really
helps you just learn a lot. Because I
think that sometimes the problem with

(02:13:23):
following tutorials or following a kind
of the stated path through school is
you're oftentimes just doing the easy
things. But then when you are up late at
night just trying to solve one particular
bug or your particular use case that

(02:13:44):
nobody else has solved before, you can't
always just look at a textbook to see how
to do that, right? You're not going to
necessarily look at a tutorial to say,
"Hey, how do I solve this brand new
problem that nobody's ever solved
before?" You just have to go and then
figure it out yourself and put different
things together. And then you're going to
become an expert in that case. And I
think that at least seeking from my own

(02:14:05):
experience, when I've tried to solve
things that I couldn't just easily
Google, I felt like my skill level just
rapidly increased as opposed to just
following tutorials. Because following
tutorials is nice, but you don't truly,
deeply understand why something is done

(02:14:25):
the way it is until you're actually the
one solving it. So yeah, I think that's
super awesome advice.(...) So now, one
thing I'm curious, is it possible if
people wanted to kind of hack on your
headsets? If they could do that, is there
some sort of developer kit? Or is this
more reserved for the people, like the

(02:14:46):
businesses who are using
your headsets directly?
Very good question. So right, actually,
what we have on our devices, given that
they run Android, a lot of things you can
do out of the box. And so we have also
properly documented SDK, say it's
probably never documented well enough.
Specifically, it's the things that are
evolving, but you might also know that.

(02:15:07):
(...) But you can do a lot of things just
with a normal device if you get one at
hand. Obviously, they're sold to B2B. You
can still get them secondhand our
devices. You can start to put your own
Android application on these devices and
start to see, hey, it's
great what I can start doing.
Sometimes you need to have system level

(02:15:27):
permissions. It's a little bit more
difficult.(...) You can always reach out
happily to us. We'll provide you certain
details that you need to get going. What
we also have when we do that with
universities from time to time, if we
have all the generations, we leave some
amounts of devices back that we can then
start using again for these kind of

(02:15:48):
tinkering things.(...) So we're doing
that together with ETH, another school in
Zurich, a very famous one, where we hand
a few hundred devices to them and they
can give it out to their scholars to
solve problems on a device. They can have
it and play around it. And so if someone
is listening in, I think you can always
try to reach out to us.(...) There's

(02:16:09):
always devices that come back from
previous users that are maybe not
refurbished. They still work. They might
have some scratches, but they can be used
again. You can start playing and build on
top of them because it's a great thing.
You have displays, you have cameras.
The audio is very well optimized for
wearing a device. It has a lot of
beamforming integrated. So you can solve

(02:16:29):
a lot of issues with it. Whereas on the
phone, it'll be more limited.
And so, yeah, for that, it is great. It
is a little bit pricey currently, at
least since they're more
focused on to the P2P side.
It's a little bit of an issue currently.
You spend quite a lot of it. If you get
one second hand, it's a lot of fun. And I
think the more and more we go now with
our software capabilities on to other

(02:16:50):
hardware devices, I guess it's not going
to take that long until you're going to
get one, which is much
cheaper to get a lot done on it.
I guess on that note, do you have any
call to action for our audience? Anything
you'd like to announce, especially
because you're opening up a new office
here in your Santa Cruz. Any open roles
for jobs either here or

(02:17:12):
remotely in other countries?
Yeah, so, yeah, thanks so much again for
the opportunity for that. I think what we
would always love, and I think that is
great, is that if everyone gets together
and solves issues, and I think you guys
do exactly that, which is also why I'm
here, right? Driving the whole community
forward to get people together to solve
issues, solve problems together and get

(02:17:34):
things done. And so I think that's the
most important part. And
that's also why we're coming here.
As I said before, right, we came, our
headquarter is currently all the way
north. It's cold. It's great weather
here. People actually love to get
together, solve problems together.
There's some of the probably the greatest
people on Earth that are all around here.
And we love to take advantage of that.

(02:17:55):
And so if there is something that we also
can help you with or anyone else who's
listening, please reach out. We'd love to
find time to actually solve things. But
we obviously also have things we are
looking for. And I think what we see is
we need a lot of app developers, way more
than we thought to get things done. If
you understand AOSP, that's also great.
It's very hard to find people.

(02:18:17):
But we sort of end up always finding the
people.(...) And I think in the end, what
is also very important is connections and
ideas. And I think that is sometimes
probably almost more valuable. I think
nowadays that I would believe the time it
takes for an idea to actually hit an end
product is probably going to be shorter,

(02:18:38):
shorter, shorter. It needs less and less
developers and less time and less money
and less effort. I think the ideas itself
become almost more valuable to the extent
that most believe ideally one would
patent ideas probably
obviously won't be possible.
But I think specifically at that point is
probably almost going to be the most
important one of what is a good idea,

(02:18:58):
what is a bad idea. And that might even
be that I have an idea that then the
majority of people would say, oh, my
gosh, that's a very bad idea.
So it's very valuable to stop before one
starts doing something. And I think we
specifically look for that if people see
and say, oh, my gosh, this is crazy.
Don't do that. And specifically, if I
look at our current, specifically Harvard

(02:19:19):
portfolio, I think there's a lot of
things that we can do better that maybe
also weren't possible back in the days.
But also thanks to feedback are now
getting better because we just really
can't see everything. And I think that's
our greatest wish that people reach out
and form a connection and really help
solving problems that are sometimes

(02:19:39):
visible and sometimes sadly not visible.
Yeah, but that sounds great. I think
that's pretty much all we had. And I
don't want to take your time. I know it's
very late there. So I
appreciate the time.
And yeah, for all those things, we'll try
to put the link in the description. So if
people want to reach out to you, they
can. But yeah, I really appreciate it.

(02:20:02):
Yeah, thank you so much. This is an
amazing conversation. Hope you got
something out of it. Hope you got
something out of your
visit here in Silicon Valley.
Definitely. As I said already, Mark, as
soon as we have our office here, we're
definitely going to invite you for some
surfing at the beach. And then we're
going to go dancing once.
Well, that sounds lovely. Yeah, I just
found out Timon does salsa too. This is
two engineers who do Latin dance, which

(02:20:26):
is very random but speaks for the
diversity and creativity in Silicon
Valley too and the
tech community at large.
But again, it was a pleasure having you.
This is an amazing conversation.
And yeah, for our
listeners,(...) see you next time.

All Episodes

The future of Augmented Reality | Interview with Real Wear CTO Timon Binder

Episode Transcript

Popular Podcasts

Bookmarked by Reese's Book Club

On Purpose with Jay Shetty

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}The future of Augmented Reality | Interview with Real Wear CTO Timon Binder